GReNaDIne: A Data-Driven Python Library to Infer Gene Regulatory Networks from Gene Expression Data.

Area of Science:

• Systems Biology
• Bioinformatics
• Computational Biology

Background:

• Inferring gene regulatory networks (GRN) from gene expression data is complex, with no single method universally outperforming others.
• Existing GRN inference tools are often implemented independently, creating challenges for users needing to compare multiple methods.
• A unified framework is needed to streamline the analysis and comparison of diverse GRN inference strategies.

Purpose of the Study:

• To develop GReNaDIne (Gene Regulatory Network Data-driven Inference), an open-source Python package for GRN inference.
• To provide a common framework integrating multiple machine learning-based GRN inference algorithms.
• To facilitate the comparison and combination of different inference methods for robust network reconstruction.

Main Methods:

• Implemented 18 distinct machine learning data-driven GRN inference algorithms.
• Included 8 generalist preprocessing techniques for RNA-seq and microarray data.
• Integrated 4 RNA-seq specific normalization techniques.
• Developed ensemble methods to combine results from multiple inference tools.

Main Results:

• GReNaDIne offers a comprehensive suite of tools for GRN inference within a single Python package.
• The package supports both RNA-seq and microarray data, with dedicated normalization for RNA-seq.
• Ensemble capabilities allow for the generation of more robust and reliable gene regulatory networks.
• Successfully validated performance using the DREAM5 challenge benchmark dataset.

Conclusions:

• GReNaDIne provides a valuable, unified toolkit for the systems biology community to infer gene regulatory networks.
• The package simplifies the process of testing and comparing various GRN inference methods.
• Its compatibility with tools like PYSCENIC enhances its utility for downstream network analysis.